Deterministic contexts
Contexts. Same sealed structure. Different institutional lens.
A context is a deterministic profile that changes lexicon, weighting, and visibility rules. The sealed structure stays invariant: same keys, same hashing, same verification — no “AI opinion”, no interpretation.
Invariant Sealed structure
Variable Vocabulary
Variable Weighting
Variable Visibility
No AI
RPO = sealed JSON core + human-readable preview + public hash. The hash proves integrity, not “truth”.
Pick a context
Institutional crisis
MEDIUM
Escalation readiness, stakeholder mapping, uncertainty logging, reconstruction under pressure.
Best when the institution needs a stable reaction path at Moment T.
Labor / litigation readiness
HIGH
Opposability framing, chain-of-custody mindset, defensible HR narrative without “AI judgment”.
Best when proceedings are plausible or already initiated.
Regulatory audit window
MEDIUM
Definition drift control, lineage discipline, repeatable evidence production over 12–36 months.
Best when compliance is a cadence, not a one-off crisis.
What changes vs what never changes
Never changes
Sealed JSON keys · hash function · verification logic · invariant flags
Changes with context
Labels · exposure vocabulary · weighting bands · what is shown in the readable preview
Deterministic contract
Same input → same output. Context only selects a profile. No probabilistic “model behavior”.
Selected context profile
Context
Institutional crisis
Default exposure band
MEDIUM
Primary intent
Escalation readiness, stakeholder mapping, uncertainty logging, reconstruction under pressure.
Lexicon shift
Institutional reaction · stakeholder exposure · crisis perimeter
Visibility rules
Operational view shows escalation grid + lineage + full hash
Weighting bias
Security logs + legal snapshot weighted higher during crisis framing
Machine-readable profile
If you later implement real configs: store these as profiles and keep pipeline folders unchanged. Context must not fork the sealed structure.